library(tidyverse)
library(ggmosaic)
library(dplyr)
breast_cancer <- read.csv("breast_cancer.csv", header=FALSE)
edited_breast_cancer = breast_cancer[2:4025, ]
age = as.numeric(edited_breast_cancer$V1)
race = edited_breast_cancer$V2
marital_status = edited_breast_cancer$V3
status = edited_breast_cancer$V16
t_stage = edited_breast_cancer$V4
n_stage = edited_breast_cancer$V5
sixth_stage = edited_breast_cancer$V6
differentiate = edited_breast_cancer$V7
grade = edited_breast_cancer$V8
a_stage = edited_breast_cancer$V9
tumor_size = as.numeric(edited_breast_cancer$V10)
estrogen_status = edited_breast_cancer$V11
progesterone_status = edited_breast_cancer$V12
regional_node_examined = as.numeric(edited_breast_cancer$V13)
regional_node_positive = as.numeric(edited_breast_cancer$V14)
survival_month = as.numeric(edited_breast_cancer$V15)

library(scales)
dead = filter(edited_breast_cancer, status == "Dead")
alive = filter(edited_breast_cancer, status == "Alive")

1. Client Bio & Recommendation

Client

Office on Women’s Health https://www.womenshealth.gov/

  • Part of the U.S. Department of Health and Human Services (DHHS)

  • It aims to promote health equity for women and girls through sex- and gender-specific approaches. OWH develops programs, educates health professionals, and disseminates health information to motivate behavior change in the public.

Recommendation

Breast cancer, as per the National Cancer Institute, is the most prevalent cancer in America. Shockingly, many cases are diagnosed in women without any identifiable risk factors other than gender and age.

Based on evidence, it’s crucial for women over 40 to undergo annual breast cancer screenings. Early detection enhances treatment success and overall health. By spreading awareness about these screenings, women can be empowered to prioritize their well-being and make informed decisions about their health. To take this initiative further, organizing free breast cancer screening events across America stands as an ideal approach to provide accessible healthcare and early detection opportunities, potentially saving many lives.

2. Evidence

The dataset, updated in November 2017 from the SEER Program of the NCI, focuses on women diagnosed with a specific type of breast cancer between 2006 and 2010. Information on patients with unknown tumor size, unchecked lymph nodes, or survival less than a month post-diagnosis was excluded. Studying 4024 patients, each row represents one breast cancer patient, and each column details various patient attributes.

Understanding Breast Cancer Risks

The boxplot indicates that breast cancer is more common among middle-aged women, placing them at a higher risk. While younger women have a lower chance of getting the disease, they still represent a portion of cases. Hence, women under 45 should stay vigilant about their health and be aware of potential risks.

ggplot(data = edited_breast_cancer, aes(x = age)) + geom_boxplot(fill = "khaki") + xlab("Age") + ggtitle("Age Distribution") + theme(axis.title.y = element_blank(), axis.text.y = element_blank(), axis.ticks.y = element_blank())

Also, according to WHO, approximately half of breast cancer cases develop in women without identifiable risk factors other than gender (female) and age (over 40). Indeed, it is shown in the following diagram.

Figure 1
dead_differentiate = dead$V7
dead_n_stage = dead$V5
dead_t_stage = dead$V4
dead_a_stage = dead$V9
dead_differentiate = ordered(dead_differentiate, levels = c("Well differentiated", "Moderately differentiated", "Poorly differentiated", "Undifferentiated"))

ggplot(data=dead, aes(x=dead_t_stage, y = after_stat(count/sum(count)))) +
geom_bar(fill = "lightsteelblue") + scale_y_continuous(labels = percent) + xlab("T Stage") + ylab("Percentage") + ggtitle("T Stage Distribution of Dead Patients")

Figure 2
ggplot(data=dead, aes(x=dead_n_stage, y = after_stat(count/sum(count)))) +
geom_bar(fill = "tan4") + scale_y_continuous(labels = percent) + xlab("N Stage") + ylab("Percentage") + ggtitle("N Stage Distribution of Dead Patients")

Figure 3
ggplot(data=dead, aes(x=dead_differentiate, y = after_stat(count/sum(count)))) +
geom_bar(fill = 'green4') + scale_y_continuous(labels = percent) + xlab("Differentiate") + ylab("Percentage") + ggtitle("Differentiate Distribution of Dead Patients")

  • T (Tumour) indicates the depth of tumour invasion - a higher number means a further spread of cancer.

  • N (Nodes) indicates lymph node involvement - a number describes the extent of cancer spreading to nodes near the bladder.

Analyzing these bar graphs, the majority of fatalities occur during the early stage. Nearly half succumbed to cancer at stage T2 and N1. Most of these patients had moderately differentiated cancer cells, indicating abnormality when compared to healthy cells.

dead_tumor_size = as.numeric(dead$V10)
ggplot(dead, aes(x = dead_tumor_size)) + geom_histogram(fill = "orchid4") + scale_x_continuous(breaks = seq(0, 140, by = 25)) + xlab("Tumor Size")  + ggtitle("Tumor Size Distribution of Dead Patients") + theme(axis.title.y = element_blank(), axis.text.y = element_blank(), axis.ticks.y = element_blank())

The data shows alarming trends, indicating most people face fatal when the tumor size was only 20-25 mm, approximately the size of a peanut. In summary, all of these graphs shows that danger can strike unexpectedly, and people might not be aware of it. Therefore, getting checked annually is crucial to catch any potential issues early.

Myths And The Truth Behind

The data challenges common beliefs about breast cancer. Some thought a small tumor or youth ensured better survival chances. However, the graph tells a different story. Even with a small tumor or at young age, the disease can be serious and progress rapidly.

Figure 1
dead_survival_month = as.numeric(dead$V15)
dead_tumor_size = as.numeric(dead$V10)
ggplot(data = dead, aes(dead_tumor_size, dead_survival_month)) + geom_point() + xlab("Tumor Size") + ylab("Survival Month") + ggtitle("Scatter Plot between Tumor Size and Survival Month of Dead Patients")

Table 1
fit = lm(dead_survival_month ~ dead_tumor_size, data = dead)
summary(fit)
## 
## Call:
## lm(formula = dead_survival_month ~ dead_tumor_size, data = dead)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -44.932 -18.420  -1.713  15.528  58.241 
## 
## Coefficients:
##                 Estimate Std. Error t value Pr(>|t|)    
## (Intercept)     48.31609    1.77078  27.285   <2e-16 ***
## dead_tumor_size -0.07285    0.04000  -1.821    0.069 .  
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 23.92 on 614 degrees of freedom
## Multiple R-squared:  0.005374,   Adjusted R-squared:  0.003754 
## F-statistic: 3.317 on 1 and 614 DF,  p-value: 0.06903

Figure 2
dead_age = as.numeric(dead$V1)
dead_survival_month = as.numeric(dead$V15)
ggplot(data = dead, aes(dead_age, dead_survival_month)) + geom_point() + xlab("Age") + ylab("Survival Month") + ggtitle("Scatter Plot between Survival Month and Age of Dead Patients")

Table 2
fit = lm(dead_survival_month ~ dead_age, data = dead)
summary(fit)
## 
## Call:
## lm(formula = dead_survival_month ~ dead_age, data = dead)
## 
## Residuals:
##     Min      1Q  Median      3Q     Max 
## -44.297 -18.803  -1.347  15.510  57.354 
## 
## Coefficients:
##             Estimate Std. Error t value Pr(>|t|)    
## (Intercept) 49.75090    5.58182   8.913   <2e-16 ***
## dead_age    -0.07508    0.09968  -0.753    0.452    
## ---
## Signif. codes:  0 '***' 0.001 '**' 0.01 '*' 0.05 '.' 0.1 ' ' 1
## 
## Residual standard error: 23.97 on 614 degrees of freedom
## Multiple R-squared:  0.000923,   Adjusted R-squared:  -0.0007042 
## F-statistic: 0.5672 on 1 and 614 DF,  p-value: 0.4516

We conducted regression tests to explore linear trends between tumor size/age and survival month of dead patients.

  • H0: There is no linear trend (Slope = 0)

  • H1: There is a linear trend (Slope != 0)

Both models showed p-value higher than 0.05, indicating no linear relation in these figures.

The tests help us figure out if these factors affect how long someone survives breast cancer. The results emphasize that regardless of age or tumor size, breast cancer can impact anyone. Catching it early through regular check-ups remains crucial for our overall health.

3. Acknowledgments

Breast cancer. (2022, August 8). Kaggle. https://www.kaggle.com/datasets/reihanenamdari/breast-cancer/data

Teng, J. (2019, January 18). SEER Breast Cancer data. IEEE DataPort. https://ieee-dataport.org/open-access/seer-breast-cancer-data

Cancer staging. (2022, October 14). National Cancer Institute. https://www.cancer.gov/about-cancer/diagnosis-staging/staging

World Health Organization: WHO & World Health Organization: WHO. (2023, July 12). Breast cancer. https://www.who.int/news-room/fact-sheets/detail/breast-cancer

Breast cancer facts and statistics 2023. (n.d.). https://www.breastcancer.org/facts-statistics

Du Cancer, C. C. S. /. S. C. (n.d.-b). Grading cancer. Canadian Cancer Society. https://cancer.ca/en/cancer-information/what-is-cancer/stage-and-grade/grading

Data1001/Data. (2020c, December 15). RGuide. https://diwarrensydney.github.io/

4. Appendix

Client Choice

Our data, sourced from the SEER Program of the NCI, is vital for cancer statistics in America. Choosing the Office on Women’s Health was a fitting decision due to the credibility of its data source and its wide recognition in America. Recognizing the financial barriers to breast cancer screenings, advocating for free screenings is crucial. By connecting with the Office on Women’s Health and SEER, our report becomes trustworthy and relevant.

Statisitcal Analyses

In the beginning, I decided to use graphs to showcase different aspects of the data. Graphs serve as powerful tools, giving a quick and clear picture of complex information, making the data more understandable for everyone. In this case, I wanted to emphasize the unexpected and abrupt nature of fatalities among cancer patients.

To show that there are no relationship between survival month and tumor size/age, linear modelling were used in second part. There are some assumptions we need to check:

  • The residuals should be independent

  • The residuals should follow a normal distribution: Using QQ Plot, generally, the values increase linearly, indicating a normal distribution. Minor deviations are observed at the extremities; however, the large sample size allows us to assume normality, as per the Central Limit Theorem.

  • The residuals should have a constant variance: Residuals Plot show random scattered around the line residuals equal to 0, which suggest linearity and homoscedasticity.

Figure 1
dead_survival_month = as.numeric(dead$V15)
dead_tumor_size = as.numeric(dead$V10)
ggplot(data = dead, aes(dead_tumor_size, dead_survival_month)) + geom_point() +geom_smooth(method = "lm", se = FALSE) + xlab("tumor_size") + ylab("survival_month")

Residual Plot
fit = lm(dead_survival_month ~ dead_tumor_size, data = dead)
plot(fit, which = 1)

QQ Plot
plot(fit, which = 2)

Figure 2
dead_age = as.numeric(dead$V1)
dead_survival_month = as.numeric(dead$V15)
ggplot(data = dead, aes(dead_age, dead_survival_month)) + geom_point() +geom_smooth(method = "lm", se = FALSE) + xlab("Age") + ylab("Survival Month")

Residual Plot
fit = lm(dead_survival_month ~ dead_age, data = dead)
plot(fit, which = 1)

QQ Plot
plot(fit, which = 2)

Limitations

The data we used is from breast cancer patients between 2006 and 2010, which was quite a long time ago. This means the numbers might have changed by now, making the information outdated and possibly not entirely accurate. This is a limitation we need to consider when looking at the data.